Fish consumption and plasma levels of organochlorines in a female population in Northern Norway

Abstract

Increased cancer incidence and mortality have been found among humans exposed to high levels of organochlorines (OCs), either accidentally or as industrial workers. In order to assess levels of OCs in Norwegian women north of the Arctic Circle and validate self-reported fish consumption as a surrogate measure of organochlorine body burden, concentrations of seven polychlorinated biphenyl (PCB) congeners [IUPAC Nos. CB-105, CB-118, CB-138 (+ CB-163), CB-153, CB-180, CB-183, CB-187], β-hexachlorocyclohexane (β-HCH), 2,2′-bis(p-chlorophenyl)-1,1-dichloroethylene (p,p′-DDE) and cis- and trans-chlordane (c-CD and t-CD) were examined in plasma samples of middle-aged women attending for health screening. Altogether, 47 of those invited (81%) completed a questionnaire and donated a suitable blood sample. The ability of questionnaire data to predict plasma levels of OCs was tested in linear and logistic regression analyses. Measured plasma concentrations were in the range reported for the general female population of other Western countries and the relative amounts of PCBs were similar to the circumpolar pattern. Intake of seagulls' eggs was a predictor of PCB congeners CB-138 (+CB-163) (p < 0.05) and CB-153 (p < 0.01). No other food category was positively associated with any compound. In contrast, duration of residence in the study municipality, body mass index (BMI) and lifetime lactation (months) were the best univariate predictors. There was an increase in β-HCH, p,p′-DDE and most of the PCBs (p < 0.05 for all) with increasing length of time a subject had lived in the municipality. BMI was a positive predictor for β-HCH (OR = 3.10, 95% CI 1.50–6.43, per 5 kg m⁻²), chlordane (OR = 2.13, 95% CI 1.12–4.05, per 5 kg m⁻²) and CB-105 and CB-153 (p < 0.05 for both). Lactation was negatively associated with all OCs (p < 0.05), except chlordane and two of the PCB congeners. Time living in the municipality and lactation explained 34% of the variance in concentration of total PCB in a multivariate model (p < 0.001). The results indicate that regular consumption of fish (mostly lean species) from the Norwegian waters is not associated with an increased body burden of OCs (e.g., of importance to cancer development), although they confirm that lactation is the most important elimination route of these contaminants in women.

---

Introduction

Sea food is rich in health-promoting nutrients, such as ω-3 fatty acids, vitamins and trace elements, and is protective against cardiovascular disease¹ and probably against cancer.² Concerns about environmental contaminants in marine organisms and their transmission to humans through the diet have prompted further studies in order to clarify the net health effect of sea food. Among the most relevant pollutants are organochlorines (OCs), which because of high fat solubility and high resistance to biodegradation, accumulate in living organisms and magnify in food chains. OCs originate from industry or use of pesticides and are spread worldwide by atmospheric and oceanic streams. In the northern hemisphere interaction of these factors, in particular, brings a burden to the Arctic ecosystem.³ Despite bans and restrictions in the use of several OCs during the last 30 years, concentrations of these contaminants in fish from sub-Arctic waters have not changed significantly over the last 15–20 years.⁴

In animal studies, many OCs are genotoxic or tumor promoters.^5–9 Increased cancer incidence and mortality have been found among humans exposed to high levels of OCs either accidentally or as industrial workers.^10,11 The International Agency for Research on Cancer (IARC) has classified the most toxic organochlorine TCDD (2,3,7,8-tetrachlorodibenzo-p-dioxin) as carcinogenic to humans, whereas others are considered as possible human carcinogens.¹² OCs with estrogenic or anti-estrogenic properties in human cell cultures have been linked to hypotheses about hormone-dependent cancer causation.^13,14 With regard to the potential role for these compounds in breast cancer etiology, epidemiological studies are inconsistent.^15–20

Uncertainties about exposure and the effect of OCs in populations consuming sea food from Norwegian waters have the potential to discredit numerous related products incorrectly, thereby indirectly depriving people of a source of important nutrients. Thus, the association between consumption of fish from Norwegian waters and the body burden of OCs is an important issue to be clarified, both from a health perspective and from a socio-economic point of view.

We conducted a study among women in Lofoten, Norway, at a latitude of 63°N, in order to assess the concentration of 13 different OCs in plasma and through a questionnaire to evaluate dietary and lifestyle factors as predictors of plasma organochlorine concentration.

Experimental

Study population

The study was conducted in the municipality of Vestvågøy, which is part of the Lofoten Islands on the north-west coast of Norway. The Lofoten Islands are well known for their fishing fleet, with a peak season in winter, when mature cod have come south from the Barents Sea to spawn. In November 1997 all women living in Vestvågøy who were born in 1955–57 were invited to participate in a health screening program organized by the Norwegian National Health Screening Service. Of the total women invited 71.6% attended. We started to include participants in our study in parallel with this screening and finished when 61 women in total had attended. All but three were invited to participate in the validation study. After receiving written information about the project, 54 of those 58 who were eligible (93%) agreed to take part. Altogether, 47 of the 58 invited women (81%) were included; four women who had incomplete questionnaire data and three whose vials containing the plasma samples were broken had to be excluded. Informed consent was obtained from all participants and the design of the study was approved by the Regional Committee for Medical Research Ethics.

Food frequency questionnaire

The semi-quantitative dietary questionnaire was restricted to the three fat-containing food categories: fish, meat and milk. In all 67 different food items were included. Thirty-four questions on habitual intake of sea food listed traditional Norwegian fish dishes. Lean fish fillets and dishes that contained mostly lean fish which was mixed with other ingredients before being cooked into fish balls, fish cakes, etc. (referred to as fish dishes in our material) were the most common fish meals in coastal areas of Norway. However, in January–March fatty liver of cod and in July–August fatty liver of saithe were served along with the lean fish fillets or as a separate dish. The sea food items were categorized into lean fish (cod, pollack, saithe, haddock), fatty fish (salmon, trout, redfish, head of redfish, herring, wolf-fish, plaice, halibut, mackerel), fish liver, roe, shellfish, whale meat, seal meat and seagulls' eggs. We asked about the consumption of lean and fatty fish fillets by season.

The questionnaire was designed to be self-instructive although assistance was offered if needed. The form was either completed at the screening center or taken home together with a stamped, addressed envelope. Reminders were not sent to those who did not return the questionnaire (n = 4; 7%).

The women were asked to record how often, on average, they had consumed each food item during the last year and to indicate the usual amount consumed on each occasion. Suggested portion sizes were given in natural or household units. With regard to fish liver, this was in terms of the number of tablespoonfuls per meal. Weights of the portions were derived from a Norwegian weights and measures table.²¹ Multiplication of the frequency of consumption by portion size and the standard portion weight gave an estimated average net weight intake of single food items per unit of time. The percentage milk fat of different dairy products given in the national table was entered as an additional factor in calculating total milk fat intake. Frequency of consumption and estimated amount eaten per unit of time were calculated both in singles and in groups of food items.

Other questionnaire variables

Apart from the main dietary section, the questionnaire secured information about age, place of birth, time living in the study municipality, body weight and height, reproductive health, breast-feeding and occupation. The women were asked to give their present weight in kilograms and height in centimeters. We obtained a body mass index (BMI) estimate for each participant by dividing the body weight in kilograms by the squared height in meters (kg m⁻²).

Blood sample analyses

Non-fasting blood samples were drawn from a cubital vein into two 7 ml Vacutainer Hemogard ethylenediaminetetraacetic acid (EDTA) glasses (Becton Dickinson, Sweden) by trained nurses. Plasma was separated by centrifugation at 2000 rev min⁻¹ for 10 min (Model 2010 centrifuge, Kubota, Tokyo, Japan) and transferred within 2 h of drawing the blood into pre-cleaned vials, which were coded and kept frozen (−20 °C) until analysis. To check for possible contamination of the glass vials, three field blanks were made from the SupraSolv solvent cyclohexane and these were fractionated, purified and analyzed in the same way as the plasma sample hexane extracts.

Selection of contaminants to be analyzed conformed with the practice used in the Arctic Monitoring and Assessment Program (AMAP).²²

The OCs measured in plasma from 47 women were seven PCB congeners [IUPAC Nos. 105, 118, 138 (+163), 153, 180, 183, 187], β-hexachlorocyclohexane (β-HCH), 2,2′-bis(p-chlorophenyl)-1,1-dichloroethylene (p,p′-DDE), cis- and trans-chlordane (c-CD and t-CD) and the toxaphenes Parlar 26 and 50. The plasma samples were extracted using liquid–liquid extraction with the sample, ethanol, de-ionized water saturated with ammonium sulfate and hexane. Internal standards were added before the first extraction. Specifically, 4 ml of plasma, to which 4 ml of ethanol and 4 ml of the de-ionized water saturated with ammonium sulfate were added, were extracted twice with 12 ml of hexane in a small glass tube. After this extraction, 90% of the lipids were removed using a gel permeation chromatography (GPC) column (105 cm × 1.0 cm id) purchased from LATEK (Eppelheim, Germany) and packed with 35 g of Biobeads S-X3. The remaining lipids were removed using small silica columns of 1.0 cm id. The silica columns were conditioned with 10 ml of hexane just before the sample was added. The following solvent combination was used as eluent for the OCs: 10 ml of hexane, 10 ml of hexane–dichloromethane (9 + 1), 10 ml of hexane–dichloromethane (4 + 6), 10 ml of dichloromethane–ethyl acetate (1 + 1). The combined fractions were evaporated to 0.5 ml using a Zymark (Hopkinton, MA, USA) Turbovap 500 closed cell concentrator, followed by a gentle flow of nitrogen for reduction to 100 µl. Gas chromatography (GC) was performed using a Fisons (Milan, Italy) 8060 Mega gas chromatograph. A 30 m × 0.25 mm id DB-5 MS column (0.25 µm film thickness) (J&W Scientific, Folsom, CA, USA) and a deactivated guard column (2.5 m × 0.53 mm id) (J&W Scientific) were used for all analyses. The gas chromatograph was further connected to a low-resolution Fisons MD 800 mass spectrometer. The internal standards, used for quantification, were C-13-labeled PCB 77, 101, 118, 144 and 178. Octachloronaphthalene (OCN) was added to calculate the recovery. The volume injected on to the GC column was 2 µl. Quantification was done using both negative chemical ionization (NCI) and positive electron ionization (EI+), both in the selected ion monitoring (SIM) mode. The different compounds were identified from their SIM masses and retention times. Peaks with differences in isotopic ratio >20% compared with the quantification standard were rejected and not quantified. For every 10 samples, a blank was analyzed to assess laboratory-derived sample contamination.

The limit of detection (LOD) was calculated using three times the area of the noise or, if peaks were found in the blanks, three times the area of the blank. The limit of quantification was set as 10 times the area of the noise or, if peaks were found in the blanks, 10 times the area of the blank.

The analytical method used in this study is based on accredited methods from the laboratory. The method was further developed in order to screen a large number of samples for a wide range of compounds while still being rapid and cost efficient. As part of the quality assurance system, the laboratory also participates in the AMAP's Human Health Inter-comparison Program for human blood samples.

Statistical analyses

The relationship between intake of fatty fish and the plasma concentration of OCs was studied by dividing the study population into three subgroups: ‘non-consumers’ who never ate fillets of fatty fish and ‘moderate consumers’ and ‘high consumers’ who had an estimated mean monthly consumption of about one and six meals of fatty fish fillets, respectively. The total PCB (ΣPCB) concentration was estimated by adding the concentration of the individual PCB congeners, whereas the sum of c-CD and t-CD gave total chlordanes (Σchlordane). Plasma concentrations below the limit of detection (LOD) were set to half the value of the LOD while observations for which the concentration of a specific compound was not determined due to interference were treated as missing in the statistical analyses. We performed correlation analyses including the generation of Pearson correlation coefficients for any association of variables suspected of interacting.

Frequency distribution patterns of the outcome variable determined the approach to analyses of variance. The normally distributed plasma concentrations of p,p′-DDE allowed linear regression models. Left-skewed plasma levels of the PCB congeners (Fig. 1) required logarithmic transformation of the dependent variables before statistical treatment. Analyses of variance were repeated with ranked independent variables, divided into tertiles. In the multiple linear regression analysis, we tested for the effect of sea food consumption after evaluating the effect of eight background variables, namely age, time living in the study municipality, height, BMI, number of children (parity), lifetime lactation, consumption of meat and consumption of milk fat. Non-significant background variables were deleted from the initial model one at a time, except for age, which we found appropriate to force into the model. The adjusted effects of intake of fish, liver, roe, shellfish and seagulls' eggs were then estimated by adding these consumption variables to the model. Residual analyses confirmed the assumptions in the model. As a result of a substantial number of observations, with plasma levels below the detection limits for β-HCH, c-CD and t-CD, these outcomes were analyzed using the cumulative ordinal logit model. With respect to logistic regression, we categorized β-HCH, c-CD and t-CD in thirds. The bottom third contained all observations with non-detectable plasma levels, while the remaining observations were split into two groups by their median. Results of the logistic regression analysis are reported as odds ratios which can be interpreted as the effect of the predictor variables on the odds of being in one higher category of plasma concentration. The toxaphenes were excluded from statistical analyses because most of the observations were below the detection limits.

Fig. 1 Profile of plasma concentrations of seven different PCB congeners among 47 women in Lofoten, Northern Norway. Subgroups may not total to 47 due to missing values. ^aMinimum plasma concentration (pg g ⁻¹): CB-105 = 11, CB-118 = 11, CB-138 (+CB-163) = 241, CB-153 = 171, CB-180 = 134, CB-183 = 18, CB-187 = 51.

An association was accepted when the 95% CI of the regression coefficient in the linear model did not include 0 or the 95% CI of the OR in the logistic model did not include 1. The calculations and statistical analyses were done with the SAS software package (SAS Institute, Version 6.12, 1996).

Results

Population characteristics

The characteristics of the study population are given in Table 1. All 47 women were born in Norway, of whom 27 (57%) were born in Vestvågøy and four (9%) in neighboring municipalities. Seventeen women (38%) had always lived in Vestvågøy and none had lived there for less than 8 years. Eight women (17%) had never given birth. Among the mothers, the mean parity was 2.9 and mean total lactation time throughout life was 21.7 months. Parity and lactation were strongly correlated (r = 0.81, p < 0.001). Self-reported weight and height from 43 study objects placed 26 women (61%) in the normal weight category (BMI = 18.5–24.9 kg m⁻²) and the rest in the overweight or obese class (BMI > 25 kg m⁻²), using the criteria of the WHO (results not shown).

Table 1 Selected characteristics among women in the Lofoten cross-sectional study, Northern Norway (n = 47)^a

Characteristic	Mean	Median	Range
a Subgroups may not total to 47 due to missing values. b Includes fish, fish products, shellfish, whale meat and seal meat.
Age/years	40.7	41	40–42
Lifetime living at Vestvågøy/years	30.7	36	8–42
Height/cm	166	167	153–176
Body mass index/kg m^−2	25.1	24.0	19–38
Parity (number of deliveries)	2.4	2.4	0–5
Lifetime lactation/months	18.0	16.0	0–70
Food consumption
All sea food^b/g week^−1	938	809	281–2829
Fatty fish/g week^−1	177	95	0–956
Lean fish/g week^−1	681	593	147–1765
Cod and saithe liver/g week^−1	2.8	1.3	0–13.1
Fish roe and caviar/g week^−1	46	25	0–350
Shellfish/week^−1	8.8	3.8	0–50
Whale meat/g week^−1	21	19	0–56
Seagulls' eggs/year^−1	0.26	0	0–2
Milk fat/g week^−1	111	92	17–261
Meat/g week^−1	795	733	150–2150

---

All participants regularly ate fish. Fish dishes were the most common meal, eaten on average 12 times a month by every woman. The mean frequency of a fish fillet meal was 11 per month, with lean and fatty fish fillets served 8.5 and 2.5 times per month, respectively. Among 10 women (21%) who never ate fatty fish fillet, five (11%) did not eat any fatty fish at all. Two women (4%) never ate fillets from white fish. The average consumption of bread with fish was 4.5 slices per week, although five participants (11%) did not eat fish in this way. Altogether 44 women (94%) ate the liver of cod or saithe served alone or with fillets, on average 3.7 times per year and a maximum of 9.1 times per year. An equal proportion of the population were whale meat eaters, with corresponding figures of 6.8 and 13 times per year. Shellfish was in the diet of 37 women (79%) and six (13%) ate two seagulls' eggs each per year.

Average consumption of meat was 18 meals per month, in addition to a slice of bread with meat daily. Every woman consumed dairy products. In fact 39 women (83%) drank milk, with the population average being one glass per day (1.5 dl) and a maximum of five glasses. Cheese was on average eaten with two slices of bread daily. Hen's eggs were eaten by 43 women (92%), with a population mean intake of 1.5 eggs per week (results not shown).

Plasma organochlorine concentrations

All 13 selected OCs were present in plasma samples from the study population (Table 2). The maximum concentration of a single compound was found for p,p′-DDE, with one observation above 5000 picograms per gram (pg g⁻¹) of plasma wet weight; another woman had plasma p,p′-DDE below the detection limit. Median p,p′-DDE was 936 pg g⁻¹ wet weight, which was on the same scale as the median of the most prominent PCB congeners. For the total number of observations, ΣPCB was the dominant organochlorine. Interference hindered the determination of the plasma concentration of the PCB congeners CB-105, CB-118, CB-180 and CB-183 in two (4%), one (2%), one (2%) and three (6%) samples, respectively. Apart from that, all the congeners were detected in all women. The relative amounts of different congeners measured by median plasma concentrations decreased in the order CB-138(+CB-163) >153 > 180 > 187 > 118 > 183 ≈ 105, with a left-skewed pattern (Fig. 1). For β-HCH, c-CD, t-CD and Parlar 26 and 50, multiple observations had plasma levels below the detection limits (54, 56, 32, 79 and 81% of observations, respectively). The order of magnitude of the remaining observations was 10–30 times smaller than the median value of p,p′-DDE or most PCBs (results not shown).

Table 2 Plasma organochlorine concentration (pg g⁻¹)^a among the study group (n = 47)^b

Characteristic	Total population (n = 47)			Fatty fish fillet intake^c
				None (n = 10)		Moderate (n = 19)		High (n = 18)
	Mean	Median	Range	Mean	Range	Mean	Range	Mean	Range
a The conversion factor to alter weight of plasma to volume of plasma is 0.9747.^51 b Subgroups may not total to 47 owing to missing values. c Divided into subgroups according to fatty fish fillet consumption during the last year: None, no consumption of fatty fish fillets; Moderate, less than two meals of fatty fish fillets per month; High, two or more meals of fatty fish fillets per month. d Includes the PCB congeners CB-105, CB-118, CB-138 (+CB-163), CB-153, CB-180, CB-183, CB-187. e Includes toxaphenes Parlar26 and 50. f Includes cis-chlordane (c-CD) and trans-chlordane (t-CD). g The value is limit of detection (LOD).
ΣPCB^d	2344	2377	772–4782	2945	1883–4782	2068	772–3336	2375	1152–4571
p,p′-DDE	1204	936	150^g–5075	1221	443–3836	1335	366–5075	1063	150^g–2355
β-HCH	75	50^g	50^g–358	79	50^g–238	72	50^g–200	77	50^g–358
ΣTox^e	128	65^g	65^g–729	148	65^g–523	151	65^g–729	94	65^g–450
ΣChlordane^f	120	46	25^g–747	195	25^g–747	123	25^g–644	84	25^g–422

---

All PCBs were positively correlated with other congeners. The strongest associations were between CB-138 (+CB-163) and CB-153 (r = 0.86), CB-183 and CB-187 (r = 0.83) and between CB-180 and CB-187 (r = 0.79). CB-180, CB-183 and CB-187 were related to all other congeners. CB-153 had the strongest correlation with ΣPCB (r = 0.89). Intra-family correlation was also observed between the chlordanes (r = 0.81; p = 0.0001 for all the noted correlation coefficients). Concentrations of p,p'-DDE, β-HCH, c-CD and t-CD were all inter-family correlated with ΣPCB (r = 0.58, r = 0.44, r = 0.36 and r = 0.43, respectively; p<0.0001 for p,p′-DDE and p = 0.05 for others) as well as with single congeners (results not shown).

For mean levels of OCs in the fatty fish consumption subgroups, there were no significant differences (t-test). Categorization of the study population by estimated net weight of the fatty fish fillets consumed per unit of time gave a very similar picture (results of t-test and alternative categorization are not shown).

Linear regression

In the univariate linear regression analysis, the regression coefficient corresponding to the age-adjusted change in p,p′-DDE level or log-transformed PCB level in picograms per gram of plasma by one unit change of the explanatory variable was examined (Table 3). The number of years lived in the municipality of Vestvågøy was a positive predictor for plasma p,p′-DDE (p = 0.02), although the number of births (p = 0.03) and lifetime lactation (p = 0.02) were associated with a reduced body burden of this OC. No consumption variable reached significance with variations in p,p′-DDE level. The time lived in the municipality was a significant predictor for the internal dose of ΣPCB and all single congeners (p < 0.05) except CB-105, CB-118 and CB-183; similarly, lifetime lactation explained variations in plasma levels of all congeners (p < 0.05) except CB-105 and CB-118. BMI was positively associated with CB-153 (p < 0.05) and CB-105 (p < 0.05), whereas parity was negatively associated with ΣPCB (p < 0.05) and CB-187 (p < 0.05; results for other than ΣPCB, CB-138 (+CB-163) and CB-153 not shown). The number of seagulls' eggs eaten per year was the only food category that explained differences in plasma concentration of PCBs. Levels of CB-138 (+ CB-163) (p < 0.05), CB-153 (p < 0.01) and ΣPCB (p < 0.05) increased with intake of eggs. With regard to the PCB congeners not included in the table, none except those mentioned above was significantly dependent on any predicting variable. Furthermore, analysis of variance with ranked independent variables, divided into tertiles, did not reveal any additional relationships.

Table 3 Results of univariate linear regression. The presented estimates are β × 10³ with 95% CI (n = 47)^a

Predictor variable^b	lnΣPCB^c/pg g^−1^d		lnCB-138 (+CB-163)/pg g^−1^d		lnCB-153/pg g^−1^d		p,p′-DDE/(pg g^−1)^d
a Some estimates may be based on fewer observations, because subjects with missing information for the actual dependent or independent variable were excluded. b Variables were age-adjusted. c Includes the PCB congeners CB-105, CB-118, CB-138 (+CB-163), CB-153, CB-180, CB-183, CB-187. d The conversion factor to alter weight of plasma to volume of plasma is 0.9747.^51 e p < 0.05. f p < 0.01. g p < 0.001.
Lifetime living at Vestvågøy/years	16.3	(5.7, 26.6)^f	15.3	(1.9, 28.8)^e	19.4	(7.3, 31.5)^f	30.3	(5.5, 55.1)^e
Height/cm	3.2	(−22.1, 28.5)	−0.54	(−28.4, 27.3)	13.8	(−11.6, 39.2)	33.6	(−14.6, 81.8)
Body mass index/kg m^−2	18.6	(−8.7, 45.9)	29.4	(−2.2, 60.9)	31.7	(3.0, 60.4)^e	38.7	(−21.2, 98.6)
Parity (number of deliveries)	−101.6	(−189.1, −14.1)^e	−69.8	(−181.7, 42.2)	−72.8	(−176.2, 30.6)	−217.0	(−411.7, −22.3)^e
Lifetime lactation/months	−13.9	(−20.5, −7.3)^g	−13.7	(−22.6, −4.8)^f	−13.2	(−21.4, −5.0)^f	−20.0	(−36.7, −3.2)^e
Food consumption
Fatty fish/g week^−1	0.081	(−0.519, 0.680)	−0.015	(−0.750, 0.720)	0.091	(−0.590, 0.773)	−0.29	(−1.6, 1.0)
Lean fish/g week^−1	0.022	(−0.322, 0.366)	−0.033	(−0.440, 0.374)	0.003	(−0.004, 0.004)	−0.04	(−0.78, 0.69)
Cod and saithe liver/g year^−1	−0.096	(−0.946, 0.754)	−0.094	(−1.110, 0.922)	−0.007	(−0.950, 0.937)	−0.53	(−2.3, 1.3)
Seagulls' eggs/year^−1	195.0	(0.23, 390.0)^e	238.0	(15.4, 460.6)^e	269.6	(68.3, 470.8)^f	145.3	(−269.4, 560.1)
Milk fat/g week^−1	−1.8	(−3.8, 0.2)	−2.2	(−4.7, 0.15)	−1.8	(−4.0, 0.5)	−1.9	(−6.4, 2.7)
Meat/g week^−1	−0.11	(−0.48, 0.25)	−0.20	(−0.63, 0.22)	−0.064	(−0.458, 0.331)	−0.11	(−0.87, 0.65)

---

In the multiple linear regression analysis with ΣPCB as the outcome variable, we obtained a model using lifetime residence and lactation, which explained 34% (p < 0.001) of the variation in plasma concentrations (results not shown). Adding age to the model did not change the estimates extensively. The model was not improved when consumption variables were added. As a result of the high correlation between parity and lactation, the effect of parity was no longer apparent in the multivariate model that included lactation.

Logistic regression

Table 4 provides the results of the logistic regression analysis with β-HCH, c-CD and t-CD as dependent variables; it gives the odds ratios (ORs) for being in one higher plasma concentration category per change in explanatory variables. We found a 63% increase in the odds of having a detectable β-HCH level in plasma for every 5 years of residence in the municipality (OR = 1.63, 95% CI: 1.17–2.28). Odds of being in one higher plasma concentration category increased, with OR = 3.10 (95% CI: 1.50–6.43) and OR = 2.13 (95% CI: 1.12–4.05) for β-HCH and Σchlordane, respectively, for every 5-unit increase in BMI. Separate analyses of the variation in concentration of the two chlordane compounds revealed a dependence on the changes in BMI; this was of the same order of magnitude as for the summary variable (results not shown). For each additional 6 months of lactation, the women reduced their odds of having a plasma concentration above the detection limits of β-HCH by 33% (OR = 0.67, 95% CI: 0.50–0.90).

Table 4 Logistic regression model for plasma β-HCH, c-CD and t-CD. Odds ratios with 95% CI for being in a higher category of plasma organochlorines (n = 47)^a

Predictor variable	β-HCH		ΣChlordane^b
a Some estimates may be based on fewer observations, because subjects with missing information for the actual dependent or independent variable were excluded. b Includes cis-chlordane (c-CD) and trans-chlordane (t-CD). c p < 0.05. d p < 0.01.
Lifetime living at Vestvågøy/5 years	1.63	(1.17, 2.28)^d	1.14	(0.88, 1.47)
Body mass index/5 kg m^−2	3.10	(1.50, 6.43)^d	2.13	(1.12, 4.05)^c
Lifetime lactation/6 months	0.67	(0.50, 0.90)^d	1.03	(0.84, 1.25)
Food consumption
Fatty fish/50 g week^−1	1.08	(0.95, 1.23)	0.90	(0.78, 1.03)
Lean fish/50 g week^−1	1.02	(0.95, 1.10)	0.99	(0.92, 1.06)
Seagulls' eggs/1 egg year^−1	1.83	(0.81, 4.13)	1.31	(0.55, 3.12)
Milk fat/20 g week^−1	0.87	(0.72, 1.06)	0.91	(0.76, 1.08)

---

Discussion

In this survey from the sub-Arctic area of Norway, all the selected OCs were present in plasma samples from the study population and every woman had measurable levels of the contaminants in her blood. Fish intake did not predict the plasma level of any of the measured OCs, however lean fish was one of the main dietary components. Consumption of seagulls' eggs was associated with an increased concentration of the PCB congeners, CB-138 (+ CB-163) and CB-153. The time spent living in this coastal area of Norway and the BMI had a positive association, whereas accumulated lactation time had a negative association, with the levels of most of the PCBs and pesticides.

All of the OCs in our study have been detected in Arctic abiotic and biotic samples.²² They have been selected for AMAP assessment because they would be expected to have biological effects on the Arctic biota if the exposures were similar to those in more polluted environments further south. According to this, the OCs determined were expected to be present in the plasma of the study population who lived close to the Arctic. Compared with the ranges of PCB concentrations in blood plasma of the samples collected from 50 wives of fishermen in Sweden in the mid-1990s, in our study the range for the congener CB-153 was lower (360–3960 versus 171–1232 pg g⁻¹ wet weight), whereas the range for CB-138 (+CB-163) was similar (210–2490 versus 241–2256 pg g⁻¹ wet weight).²³ In order to make sound comparisons with other studies, levels of CB-118, CB-138 (+CB-163), CB-153 and CB-180 should perhaps be emphasized, because they in general appear to be the four congeners with the highest concentrations. The median sum of these congeners in this study was 1875 pg g⁻¹ wet weight (result not shown), which is fairly close to the median sum of the same congeners measured in 206 Dutch women during the last month of pregnancy (2040 pg ml⁻¹) from 1990 to 1992.^24,25 In blood drawn from a group of 240 American women around 1990, the plasma concentration of DDE ranged from 140 to 39 440 pg ml⁻¹, with a mean of 7090 pg ml⁻¹, which is about six times the mean plasma p,p′-DDE concentration in our study.²⁶

In the plasma samples of our study the level of CB-153 was strongly related to ΣPCB (r = 0.89); in fact, most specific congeners were inter-related. In the Swedish study mentioned above, there was a high correlation between the plasma concentration of the sum of PCBs and CB-153, a major and very stable congener (r = 0.99).²³ This correlation was also found in an assessment of PCBs in the breast milk of 28 mothers in Oslo, Norway in 1991.²⁷ Among groups of American women, DeVoto et al.²⁸ found that blood levels of specific congeners were, in general, highly correlated. These findings support the use of CB-153 as an indicator substance when monitoring total PCB exposure and justifies measurement of a select group, rather than a large panel, of congeners in order to improve the cost-effectiveness and enhance uniformity of studies.

PCBs and p,p′-DDE comprise the bulk of OC residues found in humans, owing to their much longer half-lives in relation to other chlorinated contaminants.²⁹ The pattern of plasma concentrations found in our study clearly reflects these well-known variations in the efficiency of metabolism and excretion of different OCs. The left-skewed distribution of the PCBs in the 47 samples is a typical feature²² further enhancing the external validity of our findings.

In our study, the BMI was related to plasma concentrations of the most prominent PCB congeners and all the pesticides except p,p′-DDE. The observed associations are physiologically plausible since OCs are lipophilic compounds which enter the body through ingestion of foods with a high fat content and become stored in adipose tissue. Regarding the PCBs, similar findings have been reported by others,²⁰ yet for p,p′-DDE both positive³⁰ and negative²⁰ correlations with BMI have been found in previous studies. In a study in Germany, it was found that a high post-pregnancy BMI increased the likelihood of having a high β-HCH level and decreased the likelihood of having high PCB levels in the nursing women's milk.³¹ These findings indicate that the BMI may affect circulating levels of OCs and should also be considered as a potentially important modifying factor for exposure to lipophilic substances.

Lactation is the most important method of eliminating body stores of OCs.³² PCB levels in breast milk were inversely related to the duration of lactation in another Norwegian study in the early 1980s.³³ It is reassuring that data from our study clearly reflect an inverse association between lactation and OC body burden which is well known.^15,33,34 This justifies concern about the transmission of OCs to the breast-fed infant and about advice to pregnant and nursing women regarding the intake of potentially highly contaminated food.

It has been shown that the primary source of dietary exposure to PCBs varies with the level of food contamination and with dietary practices.⁵ Consumption of fatty fish from the Baltic Sea and the Great Lakes is clearly reflected in internal human OC doses, because of the relatively high contamination levels in aquatic organisms in these water systems.^26,35–37In certain circumpolar populations, fish, seal and beluga are the major sources of exposure.²² The observed positive association between lifetime residence in Vestvågøy and the compounds that had the longest half-lives: most of the PCBs, p,p′-DDE and β-HCH, might thus reflect long-time OC exposure and accumulation either through relatively high levels of PCBs and pesticides in locally harvested food or through specific dietary habits in Vestvågøy.

There are few published studies comparing food intake directly with plasma levels of OCs, with the exception of the evaluation of contaminated fish intake. In a German study, only modest positive correlations were observed between consumption of beef and lamb and PCBs, DDT (dichlorodiphenyltrichloroethane) and β-HCH in plasma, whereas consumption of saltwater fish had a positive correlation with PCBs.³⁸ However, plasma levels of DDE and PCBs among 240 American women were not associated with intake of meat, dairy or poultry, although consumption of fish with dark meat and eggs from two specific geographical regions were positive predictors of PCBs.²⁶

In our study, consumption of seagulls' eggs was a strong positive predictor for CB-138 (+CB-163) and CB-153, suggesting that eggs collected regionally in Lofoten, Northern Norway, may be an ongoing source of exposure to PCBs. Fish-eating birds are near the top of the food chain and tend to accumulate greater concentrations of contaminants.³⁹ In eggs collected from seabirds in Northern Norway in 1993, concentrations of the PCB congeners, CB-138 and CB-153 in particular, were high and similar to those in cod liver.⁴⁰ In a dietary survey in Northern Norway in 1998 more than a third of the population ate seagulls' eggs and the average consumption in Lofoten was 8–10 eggs per year.⁴¹ It may be that the relatively long time lapsed since the last seagulls' eggs season (e.g., May–June) contributed to a suggested underestimation by the women in our study. Consumption of seabirds' eggs among fishermen in the St. Lawrence Gulf, Canada, has indeed been shown to be strongly associated with plasma concentration of PCBs (Pearson correlation coefficient 0.27, p = 0.01).³⁹ Our results justify the established dietary guidelines from the Norwegian Food Control Authority (SNT), which warns against an annual intake of more than 5–10 seagulls' eggs.⁴¹

We did not observe positive associations between plasma levels of OCs and intake of fish, meat and dairy products. There are a number of possible explanations for this lack of dietary predictors, other than seabirds' eggs, for the body burden of OCs in our study. The reported levels of environmental contaminants in fish caught in the coastal waters of Northern Norway and the Barents Sea in the 1990s are generally low. With regard to PCBs, the concentrations are lowest in shrimps, roe and the muscle of lean fish, somewhat higher in muscle of half-fatty fish (redfish, wolf-fish, halibut) and fatty fish (herring, salmon) and highest in cod liver.⁴² Our study population ate a fish-rich diet. Nevertheless, we did not observe an association between levels of either PCBs or pesticides and intake of fish. This observation supports the questionnaire conclusion that lean fish was the major sea food consumed in this population.

Apart from fish, other fat-containing animal foodstuffs on the Norwegian food market also have low levels of PCBs.⁴³ Furthermore, the estimated exposure through diet was substantially lower in 1997 than in 1992, indicating a general decline in levels of OCs in the food supply over the last decade.³² It has been suggested that the actual levels and/or the bioavailability of OCs in foods other than fish might generally be too low to be detected in plasma.²⁶

In addition, as a result of the long half-lives of PCBs and p,p′-DDE, changes in diet over the years before exposure assessment might have masked our ability to observe dietary predictors. As a result of the high number of years that the women had lived in this coastal area and the stable availability and use of sea food in this region, however, great shifts in dietary pattern over the last few decades are not likely. A limitation of our study is the relatively small sample size, which may reduce our ability to detect weak dietary associations.

Sea food is a major contributor to the intake of ω-3 fatty acids. When the questions about sea food consumption used in this study were converted to intake of ω-3 fatty acids in a previous validation study, a significant correlation of the order of 0.55 was found with serum phospholipid ω-3 fatty acids.⁴⁴ Moreover, in a recent study in Greenland, plasma ω-3 fatty acids were strongly correlated with plasma levels of persistent organic pollutants, including 14 PCB congeners and four toxaphenes.⁴⁵ Hence the food frequency questionnaire can serve as a suitable instrument for predicting plasma levels of OCs.

Most of the participants reported eating more than one meal with fish or meat every day. This is consistent with dietary habits found in sociological studies among residents of the coastal line of Northern Norway (S. H. Eriksen, personal communication, 2001). Further, the high number of individuals classified in the upper BMI group is suggestive of the interpretation that a substantial proportion of the study population had a high daily intake of food.

The questionnaire did not cover the use of tobacco and alcohol. In a Norwegian study, there seemed to be a tendency towards higher levels of PCBs and DDE in milk samples from mothers who were smokers.³³ Smoking has a positive relationship with blood OC concentration in some studies.⁴⁵ However, Grimvall et al.²³ did not find any association between smoking habits and plasma levels of PCBs among 50 Swedish women and DeVoto et al.³⁸ found that β-HCH had a negative relationship with smoking in elderly Germans. A recent assessment of OCs in tobacco products showed that tobacco and cigarette smoke are a minor source of human exposure, in contrast to earlier studies.⁴⁶ An independent effect of alcohol consumption on OC body burden has been suggested in some studies,^45,47 which might reflect the adverse effect of alcohol on the liver's ability to metabolize OCs. Nevertheless, it is hard to believe that information about these stimulants in our study would have changed the overall results substantially.

Trained nurses ensured that samples were collected, handled and stored according to protocol. The use of field blanks controlled for inadvertent contamination of plasma samples. The blood sample collection was not uniform with regard to time of day and time since last meal, but any major variations resulting from this procedure are doubtful. One study of 31 healthy women suggested that temporal changes in OC levels within a 1–3-month period are minimal and that a single measure for estimating exposure is highly reliable for DDE and PCB.⁴⁸ Furthermore, Longnecker et al.⁴⁹ found that postprandial and fasting OC blood levels were highly correlated in 39 individuals from the general population.

As a result of the biochemical properties of OCs, plasma levels of these compounds generally correlate with the lipid profile.^23,35 The present study did not include blood lipid analyses, which circumvented expression of the results on a lipid weight basis. This limited comparisons with some published studies. Nevertheless, the work of Kuwabara et al.⁵⁰ indicates that increased concentrations of PCBs in the blood after a meal of heavily contaminated food are not associated with a corresponding change in serum lipids.

Conclusions

The study population of regular lean fish eaters appears not to be at special risk from organochlorine exposure and accumulation. The relatively low concentrations of plasma OCs observed and the lack of an association with consumption of fish indicate that worry about these contaminants should not be a deterrent for consumption of sea food from the coastal waters of Northern Norway. Similarly, our results also support the contention that the contribution of dietary OC exposure to hormone-dependent cancer causation in Norwegian women is reassuring low. However, a specific dietary habit, intake of seagulls' eggs, was strongly associated with the body burden of certain PCBs. Our results confirm that lactation is the most important elimination route for OCs in women.

The present data support the notion that the general Norwegian diet uniformly contains low levels of OCs and illustrate that in future national studies of cancer we shall have to strengthen the methods when categorizing female consumers with respect to OC exposure by use of the food frequency questionnaire alone.

Acknowledgements

This study was supported by a grant (TP 49 258) from the Norwegian Cancer Society. Our work was inspired and guided by the Arctic Monitoring and Assessment Programme.

References

1. S. Grimsgaard, K. H. Bønaa, J.-B. Hansen and E. S. P. Myhre, Am. J. Clin. Nutr., 1998, 68, 52.
2. Food, Nutrition and the Prevention of Cancer: a Global Perspective, World Cancer Research Fund in association with American Institute for Cancer Research, Washington, DC, 1997, pp. 404–416.
3. L. Ritter, K. R. Solomon and J. Forget, International Programme on Chemical Safety, Persistent Organic Pollutants, PCS/95.38, December 1995.
4. Arctic Monitoring and Assessment Programme (AMAP) Fact Sheet 1, AMAP, Oslo, October 2000.
5. International Programme on Chemical Safety, Environmental Health Criteria 140, World Health Organization, Geneva, 1993.
6. International Programme on Chemical Safety, Environmental Health Criteria 123, World Health Organization, Geneva, 1992.
7. International Programme on Chemical Safety, Environmental Health Criteria 34, World Health Organization, Geneva, 1984.
8. J. I. Goodman, D. J. Brusick, W. M. Busey, S. M. Cohen, J. C. Lamb and T. B. Starr, Toxicol. Sci., 2000, 55, 3.
9. K. Jaga and D. Brosius, Rev. Environ. Health, 1999, 14, 39.
10. P. A. Bertazzi, C. Zocchetti, S. Guercilena, D. Consonni, A. Tironi, M. T. Landi and A. C. Pesatori, Epidemiology, 1997, 8, 646.
11. A. Manz, J. Berger, J. H. Dwyer, D. Flesch-Janys, S. Nagel and H. Waltsgott, Lancet, 1991, 338, 959.
12. www.iarc.fr (11 June 2001).
13. D. T. Zava, M. Blen and G. Duwe, Environ. Health Perspect., 1997, 105, 637.
14. H.-O. Adami, L. Lipworth, L. Titus-Ernstoff, H. Chung-Cheng, A. Hanberg, U. Ahlborg, J. Baron and D. Trichopoulos, Cancer Causes Control, 1995, 6, 551.
15. I. Romieu, M. Hernandez-Avila, E. Lazcano-Ponce, J. P. Weber and É. Dewailly, Am. J. Epidemiol., 2000, 152, 363.
16. A. Demers, P. Ayotte, J. Brisson, S. Dodin, J. Robert and É. Dewailly, Cancer Epidemiol. Biomarkers Prev., 2000, 9, 161.
17. A. P. Høyer, T. Jørgensen, P. Grandjean and H. B. Hartvig, Cancer Causes Control, 2000, 11, 177.
18. E. M. Ward, P. Schulte, B. Grajewski, A. Andersen, D. G. Jr. Patterson, W. Turner, E. Jellum, J. A. Deddens, J. Friedland, N. Roeleveld, M. Walters, M. A. Butler, E. DiPietro and L. L. Needham, Cancer Epidemiol. Biomarkers Prev., 2000, 9, 1357.
19. F. Laden, G. Collman, K. Iwamoto, A. J. Alberg, G. S. Berkowitz, J. L. Freudenheim, S. E. Hankinson, K. J. Helzlsouer, T. R. Holford, H.-Y. Huang, K. B. Moysich, J. D. Tessari, M. S. Wolff, T. Zheng and D. J. Hunter, J. Natl. Cancer Inst., 2001, 93, 768.
20. M. S. Wolff, A. Zeleniuch-Jacquotte, N. Dubin and P. Toniolo, Cancer Epidemiol. Biomarkers Prev., 2000, 9, 271.
21. Measures and Weights of Food Items, National Association for Nutrition and Health, Oslo, 1989 (in Norwegian).
22. AMAP Assessment Report: Arctic Pollution Issues, Arctic Monitoring and Assessment Programme(AMAP), Oslo, 1998.
23. E. Grimvall, L. Rylander, P. Nilsson-Ehle, U. Nilsson, U. Strömberg, L. Hagmar and C. Östman, Arch. Environ. Contam. Toxicol., 1997, 32, 329.
24. S. Patandin, P. C. Dagnelie, P. G. H. Mulder, E. Op de Coul, J. E. van der Veen, N. Weisglas-Kuperus and P. J. J. Sauer, Environ. Health Perspect., 1999, 107, 45.
25. C. Koopman-Esseboom, M. Huisman, N. Weisglas-Kuperus, C. G. van der Paauw, L. G. M. Th. Tuinstra, E. R. Boersma and P. J. J. Sauer, Chemosphere, 1994, 28, 1721.
26. F. Laden, L. M. Neas, D. Spiegelman, S. E. Hankinson, W. C. Willett, K. Ireland, M. S. Wolff and D. J. Hunter, Environ. Health Perspect., 1999, 107, 75.
27. H. R. Johansen, G. Becher, A. Polder and J. U. Skaare, J. Toxicol. Environ. Health, 1994, 42, 157.
28. E. DeVoto, B. J. Fiore, R. Millikan, H. A. Anderson, L. Sheldon, W.C. Sonzogni and M. P. Longnecker, Am. J. Ind. Med., 1997, 32, 606.
29. M. P. Longnecker, W. J. Rogan and G. Lucier, Annu. Rev. Public Health, 1997, 18, 211.
30. J. M. Schildkraut, W. Demark-Wahnefried, E. DeVoto, C. Hughes, J. L. Laseter and B. Newman, Cancer Epidemiol. Biomarkers Prev., 1999, 8, 179.
31. G. Schade and B. Heinzow, Sci. Total Environ., 1998, 215, 31.
32. U. G. Ahlborg, L. Lipworth, L. Titus-Ernstoff, C.-C. Hsieh, A. Hanberg, J. Baron, D. Trichopoulos and H.-O. Adami, Crit. Rev. Toxicol., 1995, 25, 463.
33. J. U. Skaare and A. Polder, Arch. Environ. Contam. Toxicol., 1990, 19, 640.
34. L. Rylander, E. Dyremark, U. Strömberg, C. Östman and L. Hagmar, Sci. Total Environ., 1997, 207, 55.
35. E. Nieboer and G. G. Fletcher, Levels of Environmental Contaminants in Humans in the Great Lakes Basin, unpublished document, McMaster University, Hamilton, Canada, June 1994.
36. L. Asplund, B.-G. Svensson, A. Nilsson, U. Eriksson, B. Jansson, S. Jensen, U. Wideqvist and S. Skerfving, Arch. Environ. Health, 1994, 49, 477.
37. B.-G. Svensson, A. Nilsson, E. Jonsson, A. Schütz, B. Åkesson and L. Hagmar, Scand. J. Work Environ. Health, 1995, 21, 96.
38. E. DeVoto, L. Kohlmeier and W. Heeschen, Arch. Environ. Health, 1998, 53, 147.
39. É. Dewailly, C. Laliberté, L. Sauvé, L. Ferron, J. J. Ryan, S. Gingras and P. Ayotte, Chemosphere, 1992, 25, 1251.
40. R. T. Barrett, J. U. Skaare and G. W. Gabrielsen, Environ. Pollut., 1996, 92, 13.
41. www.snt.no (6 July 2001).
42. Assessment of Environmental Contaminants in Fish and Shellfish in Northern Norway, SNT-Rapport 4, Norwegian Food Control Authority (SNT), Oslo, 1997.
43. Dioxins and Dioxin-like PCBs in Food Items in Norway, SNT-Rapport 9, Norwegian Food Control Authority (SNT), Oslo, 1997.
44. A. Hjartåker, E. Lund and K. S. Bjerve, Eur. J. Clin. Nutr., 1997, 51, 736.
45. B. Deutch and J. C. Hansen, Dan. Med. Bull., 2000, 47, 132.
46. M. V. Djordjevic, J. Fan and D. Hoffmann, Carcinogenesis, 1995, 16, 2627.
47. E. F. Fitzgerald, S.-A. Hwang, B. Bush, K. Cook and P. Worswick, Am. J. Epidemiol., 1998, 148, 164.
48. M. D. Gammon, M. S. Wolff, A. I. Neugut, M. B. Terry, K. Papadopoulos, B. Levin, Q. Wang and R. M. Santella, Cancer Epidemiol. Biomarkers Prev., 1997, 6, 327.
49. M. P. Longnecker, L. Bernstein, C. L. Bird, A. K. Yancey and J. C. Peterson, Cancer Epidemiol. Biomarkers Prev., 1996, 5, 753.
50. K. Kuwabara, T. Yakushiji, I. Watanabe, S. Yoshida, K. Yoyama and N. Kunita, Bull. Environ. Contam. Toxicol., 1979, 21, 273.
51. N. W. Tietz, O. Siggaard-Andersen and E. L. Pruden, in Tietz Textbook of Clinical Chemistry, ed. C. A. Burtis and E. R. Ashwood, W. B. Saunders, Philadelphia, 2nd edn., 1994, pp. 1415–1416.

---